CN108200431A - A kind of video frequency coding rate controls frame-layer Bit distribution method - Google Patents

Abstract

A kind of video frequency coding rate control frame-layer Bit distribution method is claimed in the present invention, and the method includes：On the one hand, using frame-layer fixed weight in the sum of image information entropy, minimum transform domain absolute error and R λ model rate control algorithms, multi-stress is obtained by weighted array.Then the complexity of frame-layer image is weighed using multi-stress；On the other hand, the state of Video coding buffering area is analyzed, the bit distribution of coded frame frame-layer is adaptively adjusted in the feedback bits of buffering area obtained by calculation.It is more reasonable to be distributed the invention enables bit, can improve the precision of code check control and improve the quality of Video coding.

Description

A video coding rate control frame layer bit allocation method

technical field

The invention belongs to the field of video coding and decoding, and more specifically relates to a video coding bit rate control frame layer bit allocation method.

Background technique

70% of human information is obtained through eyes, so it is better to see it than to hear it. Human vision can directly respond to changes in the environment. With the development and application of modern information processing technology and digital storage technology, video has become one of the efficient media for transmitting, recording and reproducing information. In order to meet people's demand for high-quality video consumption, digital video has shifted from the current 720P to 1080P, and even 4K×2K. At present, 8K×4K video technology is also under development. The frame rate of digital video has been increased from 30fps to 60fps and 120fps, even towards the goal of 240fps. Since high-definition and high-frame-rate videos contain a huge amount of information, the amount of data required to describe these videos is also very large, and video traffic has shown explosive growth. Due to the existence of a large amount of spatial redundancy, temporal redundancy, statistical redundancy, visual redundancy, structural redundancy, knowledge redundancy and image redundancy in video, if prediction, transformation, quantization and entropy coding can be used to By removing these redundant information, the video data can be compressed and encoded.

Since high-performance video compression coding is one of the key technologies to alleviate the pressure of continuously growing video data on communication network transmission and storage, video coding technology has developed rapidly in recent decades, and domestic and international standardization organizations have successively formulated There are many international video coding standards. ITU-T and ISO jointly released a new generation of international video coding standard - High Efficiency Video Coding (HEVC) in January 2013. HEVC, like most video coding standards, also adopts a hybrid coding framework of prediction and transformation. Attached Figure 1 is the coding framework of the HEVC standard. Due to the introduction of new coding technologies in each video coding module, such as high-precision motion compensation, motion estimation merging, semantic-based entropy coding, adaptive loop filtering and other new technologies, the coding efficiency of HEVC is greatly improved.

In video communication, due to the limited transmission resources in the communication network, if the bit rate of the video encoder output code stream is too high, the output buffer at the sending end will overflow, causing video transmission delay or frame loss. When the receiving end decodes and plays, The video will appear intermittent or discontinuous; if the video compression ratio is high and the output bit rate of the encoding end is small, the transmission channel resources cannot be fully utilized, and the quality of the decoded video will decrease. Blocking or fuzzy edges. The main means to solve the above problems is to use bit rate control technology in the video encoder. Rate control can not only improve the utilization rate of the channel, but also ensure high video coding quality. Therefore, the rate control algorithm is of great significance to the application of video coding and normal video communication transmission, and is indispensable for current video encoders. important parts of.

The rate control process can be divided into two steps. The first step is to allocate an appropriate number of bits to the coding units of each layer, which usually includes a group of pictures (Group Of Pictures, GOP) layer, a frame layer and a basic coding unit layer. The encoder allocates an appropriate number of bits to each layer based on buffer occupancy. The second step is to achieve the pre-allocated number of bits for each layer by determining the quantization parameters at the encoding end.

A rate control algorithm with good performance can minimize the encoding distortion while accurately reaching the target bit rate, so the rate control problem is transformed into a rate-distortion optimization (RDO, Rate DistortionOptimization) as shown in formula (1) The problem is that the encoder selects the parameter that minimizes the distortion as the optimal encoding parameter when the number of encoded bits does not exceed the target number of bits. {Para} in formula (1) represents a set of coding parameters, including mode, motion information, prediction parameters, quantization parameters, etc.; λ is the Lagrangian multiplier, that is, the absolute value of the slope of the R-D (Rate-Distortion) curve.

In order to improve the flexibility of encoding, the encoder can freely choose the combination of various encoding parameters. Since the selection of different parameters will directly affect the encoding bit rate of the final video, the bit rate control algorithm is to select the appropriate encoding parameters in the encoding parameter set to achieve the target bit rate.

HEVC adopts a novel λ-domain rate control algorithm based on the R-λ model, uses the hyperbolic model to accurately describe the R-D rate-distortion model in the encoding algorithm, and calculates the compressed and encoded video by formula (2). Distortion, where R represents the compressed bit rate, represented by the bit consumed per pixel (bit per pixel, bpp); C and K are model parameters related to the content characteristics of the video sequence, different video sequences, C and K The values are different.

D(R)＝CR ^-K (2)

Rate control is based on the R-D rate-distortion model to establish the mathematical relationship between the rate R and the Lagrangian multiplier λ used for encoding, and use the method of adjusting λ to achieve the desired target rate. Formula (3) can calculate the Lagrangian multiplier λ, where α=CK, β=-K-1. The two parameters α and β are related to the content characteristics of the video sequence, and different sequences have different values. The relationship between code rate R and λ is further obtained from formula (3), as shown in formula (4).

It can be seen from formula (4) that the code rate R is completely determined by the Lagrangian multiplier λ. The schematic diagram of the relationship between λ and the R-D curve is shown in Figure 2. λ is the absolute value of the slope of the R-D curve determined by the convex envelope of all actual operating points. There is a one-to-one correspondence between the code rate R and the Lagrange multiplier λ. Since the R-D curve is a convex function, the minimization formula (1) calculated based on a certain λ value is equivalent to using a straight line whose slope absolute value is the λ value to approximate the R-D curve, and the straight line will only be tangent to the R-D curve at one point. Therefore, the λ value can directly determine the code rate R and video distortion D.

To achieve a certain assigned target code rate R, the encoder will determine the associated λ value according to equation (1) and use it in the encoding process. After the λ value used for encoding is determined, all other encoding parameters are obtained through rate-distortion optimization.

The bit allocation of HEVC code rate control is carried out at three levels: group of picture (GOP) layer, frame layer and basic coding unit layer. The target number of bits of the frame layer is allocated according to formula (5) according to the fixed weight of the frame layer in the rate control algorithm of the R-λ model.

Wherein, T _CurrPic is the number of allocated bits of the current frame, T _GOP is the number of allocated bits of the current group of pictures (GOP), Coded _GOP represents the number of bits encoded in the GOP, Indicates the frame layer fixed weight of the current image frame in the R-λ model rate control algorithm, Assign weights to the bits of all uncoded pictures in the GOP.

It can be seen from the above that since the target bits of each frame in the image group are determined in advance according to the corresponding encoding structure and encoding sequence, and are fixed weight values, the complexity of the image itself in the video sequence is not considered, and there is no Consider the state of the buffer, resulting in unreasonable allocation of bits.

Contents of the invention

The present invention aims to solve the above problems of the prior art. A reasonable bit allocation method for video coding rate control frame layer bit allocation is proposed. Technical scheme of the present invention is as follows:

A video coding rate control frame layer bit allocation method, which comprises the following specific steps:

Step 1, input a video, analyze the frame layer image texture of the video, obtain the image information entropy EI, the minimum sum of absolute errors in the transform domain SATD value;

Step 2. According to the obtained EI and SATD values, and the fixed weight of the frame layer in the R-λ model rate control algorithm Calculate the bit allocation weight ω _pic of each frame image;

Step 3, calculate the number of bits allocated by the frame layer according to formula (1);

Wherein, T _CurrPic is the number of bits allocated by the current frame; T _GOP is the number of bits allocated by the current group of pictures GOP; Coded _GOP represents the number of bits encoded in the GOP; ω _pic is the bit allocation weight of the current frame; Assign weights to the bits of all uncoded pictures in the GOP.

Step 4, analyzing the state of the video encoding buffer, calculating the feedback bit of the buffer by setting the buffer fullness;

Step 5. Calculate the number of bits to be allocated for the current frame according to the number of allocated bits T _CurrPic at the frame layer calculated in Step 3 and the feedback bits Δ _T of the buffer.

Further, the image information entropy EI in the step 1 is calculated by formula (2).

Among them, p(x) is the probability of occurrence of image x gray level, and N is the maximum gray level of the image.

Further, the SATD value of the minimum sum of absolute errors in transform domain in step 1 is calculated by formula (3).

Wherein, M is the number of pixels in the row and column of the pixel block, and h _{i, j} are the corresponding values of the pixel block after Hadamard transformation.

Further, the frame layer bit allocation weight ω _pic in the step 2 is calculated by formula (4).

Among them, EI _i is the information entropy of the current frame image; SATD _i is the sum of the minimum transform domain absolute errors of the current frame image; Indicates that the current frame image has a fixed weight at the frame layer in the R-λ model rate control algorithm; a and b are weighting coefficients, and the values of these two coefficients are greater than 0 and less than 1.

Further, the feedback bit calculation process of the buffer in the step 4 is as follows:

Calculate the target number of bits allocated by the coded image frame, calculate the number of bits consumed in the actual encoding process of the coded image frame, find the absolute difference sum of the two bits, and then calculate the rest of the buffer by formula (5) number of bits.

Among them, T _bufleft is the number of remaining bits in the buffer; i represents the sequence number of the encoded frame; n is the sequence number of the current frame to be encoded; is the number of bits allocated to the encoded picture frame in the current picture group GOP; is the actual number of bits consumed by the encoded image in the current image group; in order to avoid buffer overflow, before encoding an image frame, first set a target buffer level L, so that the buffer fullness after encoding is as close as possible to this value, the target The calculation formula of the buffer level L is shown in (6).

L=μ×B _d (6)

_Bd = R/f (7)

Among them, the coefficient μ is a certain value between 0 and 1; B _d is the size of the buffer; R is the channel rate, which can be pre-set in the configuration file; f is the frame rate.

The formula for calculating the feedback bits of the buffer is shown in (8).

_ΔT ＝η×(LT _bufleft ) (8)

Among them, Δ _T represents the feedback bit of the buffer; the coefficient η is a value between 0 and 1; L is the target buffer level; T _bufleft represents the number of remaining bits in the buffer.

The number of bits T _CurrPicbuf allocated based on the buffer fullness is calculated by formula (9).

Further, the step 5 obtains the final frame layer bit allocation formula as shown in (10) by adopting the allocation weight based on the comprehensive factor and the buffer feedback bit weighted average method.

T _CurrPicfal = γ×T _CurrPic + (1-γ)×T _CurrPicbuf (10)

Among them, T _CurrPicfal represents the number of bits finally allocated to the current frame; T _CurrPic is the number of bits allocated for the current frame according to the allocation weight based on the comprehensive factor; T _CurrPicbuf is the number of bits allocated based on the buffer fullness; γ is the weighting coefficient, Its value ranges from 0 to 1, and the value is selected according to different configuration files.

Combine formulas (1), (9) and (10) to obtain the calculation formula of T _CurrPicfal as shown in (11).

Advantage of the present invention and beneficial effect are as follows:

The invention solves the problem that the code rate control algorithm in the current video encoding does not consider the complexity of the video sequence image itself, nor the state of the encoder buffer area, which leads to unreasonable bit allocation. The present invention calculates the sum of the image information entropy and the absolute error of the minimum transformation domain, combined with the fixed weight of the frame layer in the R-λ model code rate control algorithm, constructs the distribution weight based on the comprehensive factor, and calculates the buffer by analyzing the fullness of the buffer The feedback bits of the area, deduce the calculation formula of the final allocated bits of the current frame. By adopting the frame layer bit allocation method proposed by the present invention, the frame layer bit allocation can be improved, the rationality of bit allocation can be improved, and the accuracy of code rate control can be improved.

Description of drawings

FIG. 1 is a schematic diagram of a coding framework of the video coding standard HEVC according to an embodiment of the present invention;

Figure 2 is a schematic diagram of the relationship between the Lagrange multiplier λ and the R-D curve;

FIG. 3 is a flow chart of a video coding rate control frame layer bit allocation method in an embodiment of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be described clearly and in detail below with reference to the drawings in the embodiments of the present invention. The described embodiments are only some of the embodiments of the invention.

The technical scheme that the present invention solves the above problems is:

Step 1. Analyze the image texture of the frame layer to obtain the image information entropy EI (Entropy Information) and the minimum sum of absolute errors in the transform domain SATD (Sum of Absolute Transformed Difference) value;

Step 2. According to the obtained EI and SATD, and the fixed weight of the frame layer in the R-λ model rate control algorithm Calculate the bit allocation weight ω _pic of each frame image;

Step 3: Calculate the number of bits allocated by the frame layer according to the following formula (1).

Wherein, T _CurrPic is the number of allocated bits of the current frame; T _GOP is the number of allocated bits of the current group of pictures (GOP); Coded _GOP represents the number of bits encoded in the GOP; ω _pic is the bit allocation weight of the current frame; Assign weights to the bits of all uncoded pictures in the GOP.

Step 4: Calculate the feedback bits of the buffer by analyzing the buffer status of the video encoding end and setting the fullness of the buffer.

According to formula (2) and formula (3), the image information entropy EI and the minimum sum of absolute errors in transform domain SATD are calculated respectively. According to EI, SATD and the current frame image, the frame layer fixed weight in the R-λ model rate control algorithm Formula (4) is used to obtain the distribution weight ω _pic based on the comprehensive factor.

Among them, p(x) is the probability of the gray level of image x; N is the maximum gray level; M is the number of pixels in the rows and columns of the pixel block; h _{i, j} are the corresponding values of the pixel block after Hadamard transformation; a , b is a weighting coefficient, and its value range is between 0 and 1.

The feedback bit Δ _T of the buffer is calculated according to formulas (5)-(8).

B _d =R/f (5)

L=μ×B _d (6)

_ΔT ＝η×(LT _bufleft ) (8)

Among them, B _d is the size of the buffer; R is the channel rate, its value can be set in the configuration file in advance; f is the frame rate; Δ _T represents the feedback bit of the buffer; 1; L is the target buffer level; T _bufleft indicates the amount of remaining data in the buffer; T _bufleft is the amount of remaining data in the buffer; i indicates the serial number of the encoded frame; n is the serial number of the current encoded frame; is the number of bits allocated for the encoded image frame in the current group of pictures (GOP); is the actual number of bits consumed for encoded image frames in the current image group.

The final number of frame layer bits is calculated according to the following formula (9).

Among them, T _CurrPicfal represents the number of bits that the current frame is finally allocated; T _GOP is the number of allocated bits of the current GOP; Coded _GOP represents the number of encoded bits in the GOP; ω _pic is the bit allocation weight of the current frame; Assign weights to the bits of all uncoded images in the GOP; R is the channel rate, its value can be set in the configuration file in advance, f is the frame rate; Δ _T is the feedback bit of the buffer; γ is the weighting coefficient, and its value range Between 0 and 1, select its value according to different configuration files.

Example:

A video coding rate control frame layer bit allocation method based on comprehensive factors and buffer states. In the process of video coding rate control, the method aims at texture features of different video sequences by calculating image information entropy and minimum transform domain The sum of absolute errors, combined with the fixed weight of the frame layer in the R-λ model rate control algorithm, constructs the distribution weight based on the comprehensive factor, and considers the buffer fullness of video coding, and derives the final number of bits allocated for the current frame. calculation formula. The present invention can improve the rationality of frame layer bit allocation, improve the accuracy of code rate control, and improve rate-distortion performance.

The above embodiments should be understood as only for illustrating the present invention but not for limiting the protection scope of the present invention. After reading the contents of the present invention, skilled persons can make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims (6)

1. a video coding rate control frame layer bit allocation method, is characterized in that, specifically comprises the following steps: Step 1, input a video, analyze the frame layer image texture of the video, obtain the image information entropy EI, the minimum sum of absolute errors in the transform domain SATD value; Step 2. According to the obtained EI and SATD values, and the fixed weight of the frame layer in the R-λ model rate control algorithm Calculate the bit allocation weight ω _pic of each frame image; Step 3, calculate the number of bits allocated by the frame layer according to formula (1); Wherein, T _CurrPic is the number of bits allocated by the current frame; T _GOP is the number of bits allocated by the current group of pictures GOP; Coded _GOP represents the number of bits encoded in the GOP; ω _pic is the bit allocation weight of the current frame; Assign weights to the bits of all uncoded pictures in the GOP; Step 4, analyzing the state of the video encoding buffer, calculating the feedback bit of the buffer by setting the buffer fullness; Step 5. Calculate the number of bits to be allocated for the current frame according to the number of allocated bits T _CurrPic at the frame layer calculated in Step 3 and the feedback bits Δ _T of the buffer. 2. A kind of video coding rate control frame layer bit allocation method according to claim 1, characterized in that, the image information entropy EI in the step 1 is calculated by formula (2). Among them, p(x) is the probability of occurrence of image x gray level, and N is the maximum gray level of the image. 3. A kind of video coding rate control frame layer bit allocation method according to claim 1, characterized in that the minimum transform domain absolute error sum SATD value in step 1 is calculated by formula (3). Wherein, M is the number of pixels in the row and column of the pixel block, and h _{i, j} are the corresponding values of the pixel block after Hadamard transformation. 4. a kind of video coding rate control frame layer bit distribution method according to claim 1, is characterized in that, the frame layer bit distribution weight ω _pic in the described step 2 is calculated by formula (4); Among them, EI _i is the information entropy of the current frame image; SATD _i is the sum of the minimum transform domain absolute errors of the current frame image; Indicates that the current frame image has a fixed weight at the frame layer in the R-λ model rate control algorithm; a and b are weighting coefficients, and the values of these two coefficients are greater than 0 and less than 1. 5. a kind of video encoding code rate control frame layer bit allocation method according to claim 1, is characterized in that, the feedback bit calculation process of buffer zone in the described step 4 is as follows: Calculate the target number of bits allocated by the coded image frame, calculate the number of bits consumed in the actual encoding process of the coded image frame, find the absolute difference sum of the two bits, and then calculate the rest of the buffer by formula (5) number of bits; Among them, T _bufleft is the number of remaining bits in the buffer; i represents the sequence number of the encoded frame; n is the sequence number of the current frame to be encoded; is the number of bits allocated to the encoded picture frame in the current picture group GOP; is the actual number of bits consumed by the encoded image in the current image group; in order to avoid buffer overflow, before encoding an image frame, first set a target buffer level L, so that the buffer fullness after encoding is as close as possible to this value, the target The calculation formula of the buffer level L is shown in (6); L=μ×B _d (6) _Bd = R/f (7) Among them, the coefficient μ is a certain value between 0 and 1; B _d is the size of the buffer; R is the channel rate, which can be set in the configuration file in advance; f is the frame rate; The calculation formula of the feedback bit of the buffer is shown in (8) _ΔT ＝η×(LT _bufleft ) (8) Among them, Δ _T represents the feedback bit of the buffer; the coefficient η is a certain value between 0 and 1; L is the target buffer level; T _bufleft represents the number of remaining bits in the buffer; The number of bits T _CurrPicbuf allocated based on the buffer fullness is calculated by formula (9). . 6. a kind of video encoding code rate control frame layer bit distribution method according to claim 5, it is characterized in that, described step 5 obtains final The allocation formula of frame layer bits is shown in (10); T _CurrPicfal = γ×T _CurrPic + (1-γ)×T _CurrPicbuf (10) Among them, T _CurrPicfal represents the number of bits finally allocated to the current frame; T _CurrPic is the number of bits allocated for the current frame according to the allocation weight based on the comprehensive factor; T _CurrPicbuf is the number of bits allocated based on the buffer fullness; γ is the weighting coefficient, Its value ranges from 0 to 1, and the value is selected according to different configuration files; Simultaneous formula (1), (9) and (10), obtain the computing formula of T _CurrPicfal as shown in (11); .